Very-long-chain Acyl-CoA Synthetases*

As early as 1953, it was recognized that there were at least three enzymes with ACS activity, activating short-chain (2–3 carbons), medium-chain (4–12 carbons), and long-chain ( 12 carbons) fatty acid substrates (1). The need for enzymes capable of activating saturated verylong-chain fatty acids (VLCFAs; containing 22 carbons) became evident with the discovery of elevated VLCFA levels in X-linked adrenoleukodystrophy (X-ALD) and other peroxisomal disorders (2). Formation of VLCFA-CoA is essential for degrading excess VLCFA via peroxisomal -oxidation. Singh et al. (3) were the first to demonstrate that rat brain microsomes could activate the VLCFA lignoceric acid (C24:0). Although purification of an enzyme with very-long-chain ACS (ACSVL) activity from mammalian tissues proved to be challenging, Hashimoto and co-workers (4) were successful. By following C24:0 activation, they achieved a 479-fold purification of a 70-kDa protein from rat liver peroxisomes. cDNA encoding this enzyme was cloned and expressed in COS-1 cells, yielding an enzymatically active protein (5). The endogenous enzyme localized to peroxisomes andmicrosomes and was found only in liver and kidney; thus, earlier reports of ACSVL activity in brain suggested the existence of additional enzyme(s) (3). Two years prior to the purification and cloning of ACSVL, Schaffer and Lodish (6) independently cloned cDNA encoding a novel protein that they called “fatty acid transport protein” (FATP). FATP was identified, along with a long-chain ACS, when a cDNA librarywas screened for proteins that augmented uptake of a fluorescent fatty acid, C1-BODIPY-C12, by 3T3-L1 adipocytes. mRNA encoding this 63-kDa protein was also expressed in heart, skeletal muscle, brain, and kidney. When overexpressed in 3T3 cells, FATP was associated with the plasmamembrane fraction and stimulated the uptake of radiolabeled palmitic (C16:0), oleic (C18:1), and arachidonic (C20:4) acids, thereby displaying many of the properties expected for a transporter. When the ACSVL amino acid sequence of Hashimoto and co-workers (5) was compared with data base sequences, the protein it was most homologous to was not, as expected, long-chain or other ACSs, but rather FATP. Thus began a controversy as to whether FATP and related proteins are ACSs, fatty acid transporters, or both; this controversy remains, in part, unresolved.

Activation of fatty acids by thioesterification to coenzyme A is a fundamental metabolic process that can be found in all organisms from Archaea to man (1). Formation of acyl-CoA allows an otherwise non-reactive fatty acid to participate in biosynthetic or catabolic pathways. This two-step reaction is catalyzed by acyl-CoA synthetase (ACS 2 ; EC 6.2.1.x).
The need for enzymes capable of activating saturated verylong-chain fatty acids (VLCFAs; containing Ն22 carbons) became evident with the discovery of elevated VLCFA levels in X-linked adrenoleukodystrophy (X-ALD) and other peroxisomal disorders (2). Formation of VLCFA-CoA is essential for degrading excess VLCFA via peroxisomal ␤-oxidation. Singh et al. (3) were the first to demonstrate that rat brain microsomes could activate the VLCFA lignoceric acid (C24:0).
Although purification of an enzyme with very-long-chain ACS (ACSVL) activity from mammalian tissues proved to be challenging, Hashimoto and co-workers (4) were successful. By following C24:0 activation, they achieved a 479-fold purification of a 70-kDa protein from rat liver peroxisomes. cDNA encoding this enzyme was cloned and expressed in COS-1 cells, yielding an enzymatically active protein (5). The endogenous enzyme localized to peroxisomes and microsomes and was found only in liver and kidney; thus, earlier reports of ACSVL activity in brain suggested the existence of additional enzyme(s) (3).
Two years prior to the purification and cloning of ACSVL, Schaffer and Lodish (6) independently cloned cDNA encoding a novel protein that they called "fatty acid transport protein" (FATP). FATP was identified, along with a long-chain ACS, when a cDNA library was screened for proteins that augmented uptake of a fluorescent fatty acid, C 1 -BODIPY-C 12 , by 3T3-L1 adipocytes. mRNA encoding this 63-kDa protein was also expressed in heart, skeletal muscle, brain, and kidney. When overexpressed in 3T3 cells, FATP was associated with the plasma membrane fraction and stimulated the uptake of radiolabeled palmitic (C16:0), oleic (C18:1), and arachidonic (C20:4) acids, thereby displaying many of the properties expected for a transporter. When the ACSVL amino acid sequence of Hashimoto and co-workers (5) was compared with data base sequences, the protein it was most homologous to was not, as expected, long-chain or other ACSs, but rather FATP. Thus began a controversy as to whether FATP and related proteins are ACSs, fatty acid transporters, or both; this controversy remains, in part, unresolved.

The ACSVL Enzyme Family
Proteins orthologous and paralogous to ACSVL and FATP have been identified in many species. The ACSVL 3 family in humans and rodents contains six proteins (Table 1) (7)(8)(9). Zebrafish (Danio rerio), fruit fly (Drosophila melanogaster), and roundworm (Caenorhabditis elegans) ACSVL families have eight, six, and two proteins, respectively (10), whereas a yeast (Saccharomyces cerevisiae) and a bacterium (Mycobacterium tuberculosis) have only a single homologous protein (7,11). The six human ACSVL proteins (containing 619 -730 amino acids) share 37-59% identity but are only 17-25% identical to other human ACSs (10). All enzymatically active ACSs have two conserved amino acid sequence domains: a 10-residue AMP-binding domain (Motif I) that is highly conserved from bacteria to man (9) and a recently revised 35-residue domain (Motif II) containing a sequence originally proposed as a fatty acid-binding "signature motif" (10). Motif II has been used to assign ACSs to subfamilies (9).
Examination of the crystal structure of Thermus thermophilus long-chain ACS suggests that Motif II may not be involved in chain length recognition. Its fatty acid-binding tunnel is formed by four separate smaller domains located between Motifs I and II (12). A conserved arginine residue at the center of Motif II interacts with the 3Ј-hydroxyl group of AMP and not the fatty acid (10,13). Although this motif may not dictate fatty acid substrate specificity, enzymes grouped into subfamilies of short-, medium-, and long-chain enzymes using Motif II seem to prefer these substrates (1,10). In contrast, the enzymes grouped as ACS-VLs have all been shown to be capable of activating C24:0, which is a poor substrate for other ACSs. Inspection of human ACS regions homologous to domains forming the T. thermophilus ACS fatty acid-binding pocket did not suggest that they contain features useful for predicting substrate specificity. 4 At present, therefore, substrate utilization must be established experimentally.
Like other ACSs, the ACSVL proteins differ in their tissue expression patterns, subcellular locations, and substrate specificities, suggesting that each plays a unique role in lipid metabolism. This minireview will focus on mammalian ACSVLs. ACSVL1-ACSVL1 (also known as VLCS, VLACS, and FATP2) ( Table 1) is the enzyme purified by Hashimoto and coworkers (5). It is expressed mainly in liver and kidney and has been localized to both endoplasmic reticulum membranes and peroxisomes (14). When overexpressed in COS-1 cells, human ACSVL1 activates C16:0 and C24:0 and the branched-chain fatty acids phytanic and pristanic acids. In peroxisomes, ACSVL1 is oriented facing the matrix, suggesting that it could activate pristanic acid, generated by phytanic acid ␣-oxidation, to its CoA derivative for further degradation via peroxisomal ␤-oxidation. Although the level of ACSVL1 in microsomes is almost 10-fold greater than that in peroxisomes, its microsomal function remains obscure. Trihydroxycholestanoic acid (THCA) is also a substrate for ACSVL1, implicating this enzyme in the de novo synthesis of bile acids from cholesterol (15). However, it is unclear whether this process takes place in peroxisomes or microsomes (Fig. 1).
An ACSVL1 knock-out mouse was produced that was viable, fertile, and exhibited no obvious phenotypic abnormalities. Liver and kidney peroxisomes from this mouse had reduced rates of VLCFA ␤-oxidation, yet plasma and tissue VLCFA levels were normal (16). Although decreased peroxisomal ACSVL activity was reported in fibroblasts from patients with the neurodegenerative disease X-ALD, no changes in ACSVL1 expression were found either in these cells or in X-ALD mouse tissues (17).
ACSVL2-Human ACSVL2 (VLCS-H1 and FATP6) ( Table  1) cDNA has been cloned previously (8,18). ACSVL2 expression was limited to the heart on Northern blots (18). However, expressed sequence tags from human heart, brain, lung, larynx, and ovary libraries have been identified, and the protein has been detected in hair follicle epithelial cells (19). Gimeno et al. (18) localized the protein to the sarcolemma of cardiac myocytes in mice and rhesus monkeys. Although ACS activity of human ACSVL2 has not been demonstrated, mouse ACSVL2 expressed in a yeast model system activated C18:1, C20:4, and C24:0 (20).
ACSVL3-ACSVL3 (FATP3) is the most recently characterized member of the ACSVL family (21). Mouse ACSVL3 mRNA was found primarily in steroidogenic tissues (adrenal, testis, and ovary) and in brain, and the protein localized to punctate subcellular vesicles that sedimented with mitochondria (21). Loss-offunction studies in RNA interference-treated mouse MA-10 Leydig cells demonstrated that C16:0 and C24:0 are ACSVL3 substrates and that this enzyme accounts for 30 and 26% of the total cellular activation of these fatty acids, respectively. FATP1-More is currently known about FATP1 (proposed designation ACSVL4), the protein originally called FATP (6), than for the other members of the ACSVL family. FATP1 mRNA is highly expressed in muscle, heart, brain, and adipose tissue (6). When purified to homogeneity following overexpression in COS-1 cells, FATP1 catalyzed the activation of C16:0 and C24:0 (22). Increased incorporation of fatty acids into triacylglycerol was observed when FATP1 was overexpressed in primary human myocytes or HEK293 cells (23,24). In myocytes, decreased oxidation of C16:0 or C18:1 to CO 2 was also noted. In 293 cells, FATP1 overexpression was accompanied by an increase in 1,2-diacylglycerol acyltransferase activity. Because fatty acid incorporation into glycerophospholipids and sphingomyelin was unaffected by FATP1 overexpression, it was suggested that this protein channels exogenous fatty acids into triacylglycerol biosynthesis. Reduced triacylglycerol deposition in 3T3-L1 adipocytes with short hairpin RNA-induced FATP1 knockdown supports this conclusion (25). Interestingly, lack of FATP1 did not otherwise affect preadipocyte differentiation.
FATP1 expression was increased during adipocyte differentiation in culture both in 3T3-L1 cells and in HIB-1B brown adipocytes (6,26). Insulin had pleiotropic effects on FATP1 expression, with early repression followed by robust stimulation during differentiation and subsequent repression in mature adipocytes (27). Ligands that activate PPAR␣ and PPAR␥ increased FATP1 expression, but in a tissue-specific manner (28). The ␤-adrenergic agonist isoproterenol increased FATP1 protein levels in HIB-1B cells (26), whereas tumor necrosis factor-␣ reduced FATP1 expression in 3T3-L1 adipocytes (29). In brown fat, FATP1 expression was increased by cold exposure, consistent with a role in non-shivering thermogenesis (26).
Originally reported to be in the plasma membrane of 3T3-L1 cells, subsequent studies indicated that in adipocytes and skeletal muscle cells, FATP1 translocates to the cell surface from an intracellular perinuclear compartment upon insulin stimulation (29,30). This phenomenon may be tissue-specific, as it was not observed in cardiomyocytes (31).
Topologic studies showed that FATP1 has at least one transmembrane-spanning domain and several membrane-associated domains, with the N terminus oriented toward the extracellular space and the C terminus facing the cytoplasm (32). Co-immunoprecipitation studies suggested that FATP1 exists as a homodimer that also interacts with the long-chain ACS ACSL1 (33,34). The interaction between FATP1 and ACSL1 was not affected by acute insulin or isoproterenol treatment.
A transgenic mouse with heart-specific FATP1 overexpression showed evidence of increased cardiac uptake of labeled C16:0 that was mainly in the free fatty acid pool (35). By 3 months of age, these mice developed a lipotoxic cardiomyopathy characterized by electrocardiogram abnormalities, impaired ventricular function, and atrial enlargement similar to that seen in diabetes. FATP1 knockout mice fed a normal chow diet were phenotypically normal, with no cardiac abnormalities (31,36). However, when fed a high fat diet, these mice were protected from fat-induced insulin resistance in skeletal muscle. FATP1-null mice had smaller lipid droplets in brown adipose tissue and were unable to maintain body temperature during cold exposure. A link between FATP1 and obesity and type 2 diabetes in humans has been proposed. A polymorphism in intron 8 of the human FATP1 gene is associated with elevated triacylglycerol levels and alterations in low density lipoprotein particle size distribution (37,38). However, a study of lean versus obese non-diabetic subjects reported a gender-related difference in skeletal muscle FATP1 expression, but did not support the conclusion that this protein contributes significantly to obesity or type 2 diabetes (39). A study of monozygotic twins with concordant versus discordant body mass indices found that FATP1 levels correlated somewhat with acquired differences in free fatty acid levels, but that FATP4 (see below) was more likely to play a role in acquired obesity (40).
Stahl et al. (44) reported that FATP4 was located in the apical membrane of mature mouse enterocytes and suggested that a major function of this protein was intestinal fatty acid uptake. However, the primary defect in FATP4null mice was in the skin, where hyperproliferative hyperkeratosis, defective barrier function, and restrictive dermopathy resulted in death shortly after birth or embryonic lethality (36,45). FATP4 is robustly expressed in human and mouse skin sebaceous glands and is found at lower levels in keratinocytes (19). The epidermis of FATP4-null mice had decreased levels of phosphatidylcholine, phosphatidylethanolamine, and cholesterol esters and increased ceramides. In addition, the ceramide fatty acid composition was altered, with increased long-chain and decreased very-long-chain saturated and hydroxy fatty acids (45). Skin fibroblasts from FATP4-deficient mice activated C24:0 at rates Ͻ20% of those in normal cells, suggesting that this enzyme is the principal ACSVL in this cell type (46). Peroxisomal VLCFA degradation, as well as incorporation of C24:0 into triacylglycerol, major glycerophospholipid species, and cholesterol esters, was reduced by Ͼ50% in FATP4-deficient cells, indicating that the enzyme is present in multiple subcellular locations. This was confirmed by immunofluorescence and Western blot studies showing that the protein associated with peroxisomes, mitochondrial/mitochondrion-associated membranes, and the endoplasmic reticulum, but not with the plasma membrane. Although initial studies suggested a plasma membrane location for FATP4 in adipocytes and intestinal epithelial cells (29,44), other studies suggested that the protein localized to intracellular membranes in these same cell types (25,47). The nature of these discrepancies remains unresolved.
Like FATP1, FATP4 levels were increased in 3T3-L1 cells upon differentiation to adipocytes, yet RNA interference-mediated FATP4 knockdown did not affect preadipocyte differentiation or FATP1 levels (25,29). Unlike FATP1, FATP4 was not recruited to the plasma membrane by insulin. 3T3-L1 adipocytes with FATP4 knockdown had reduced triacylglycerol accumulation, in agreement with the fibroblast results (25). PPAR␥ has been implicated in the regulation of FATP4 levels, as knockdown in adipocytes led to increased PPAR␥ levels, and pharmacological activation of PPAR␥ plus retinoid X receptor in primary human trophoblastic cells increased expression of FATP4 (25, 48). A role for FATP4 in obesity and insulin resistance syndrome has been proposed on the basis of human studies. In a study of Finnish twins, FATP4 levels were found to be up-regulated in obese subjects independently of genetic background (40). In a study of Swedish men, heterozygosity for an infrequent polymorphism (G209S; allele frequency 0.05) was associated with a lower body mass index and lower blood triacylglycerol and insulin levels (49).
ACSVL6-ACSVL6 (ACSB, BACS, VLCS-H2, VLACSR, and FATP5) is a bile acid-CoA ligase that was originally described as a liver-specific homolog of ACSVL1 (50). ACSVL6 expression is limited to liver (50). When overexpressed in COS-1 cells, ACSVL6 protein was found primarily in endoplasmic reticulum membranes, with both the N and C termini oriented toward the cytoplasm. In contrast, the endogenous protein was detected in the basal plasma membrane of hepatocytes, adjacent to sinusoids (51). Although the overexpressed enzyme weakly activated C24:0, its preferred substrates are primary (cholic and chenodeoxycholic acids) and secondary (deoxycholic and lithocholic acids) bile acids and the bile acid precursor THCA (15). Formation of glycine or taurine conjugates requires bile acyl-CoA, and thus, a role for ACSVL6 in reconjugation of bile acids returning to the liver via the enterohepatic circulation was proposed.
Investigation of an ACSVL6 knock-out mouse revealed defects in both bile acid and fatty acid metabolism (51,52). Hepatocyte triacylglycerol and free fatty acid levels were reduced in knock-out mice. Although total bile acid concentrations in liver and bile were unchanged, bile acids in the gallbladder were mainly unconjugated. The small amount of conjugated bile acids were primary, not secondary, confirming that the main function of ACSVL6 is in bile acid recycling and not in de novo synthesis. Despite bile acid metabolism abnormalities, these mice had relatively normal fat absorption, but failed to gain weight on a high fat diet because of increased energy expenditure and decreased food intake.

ACSVL Enzymes and Fatty Acid Transport
The ACS activity of ACSVL family proteins is now firmly established. Nonetheless, all six members of this family have also been cloned, characterized, and referred to as FATPs. Two schools of thought exist regarding the mechanism by which cells take up fatty acids. The biophysical model argues that membrane adsorption and passive diffusion via "flip-flop" of fatty acids across the membrane bilayer proceed rapidly enough to support cellular needs, even in rapidly metabolizing tissues such as heart and muscle (53). Desorption from the bilayer represents the rate-limiting step in this process. In contrast, others have proposed the need for fatty acid membrane transporters, including fatty acid translocase, plasma membrane fatty acidbinding protein, and members of the FATP family (54).
Experimental resolution of this controversy has been hampered by difficulties in distinguishing between pure transport and transport plus subsequent metabolism. The membrane topography of FATP1 does not resemble that of conventional transporters (32), and localization of endogenous FATPs to the plasma membrane is not consistently observed. Metabolic trapping of fatty acids via the ACS activity of FATPs has been pro-posed as one mechanism by which these proteins facilitate fatty acid uptake. As fatty acids diffuse by flip-flop across the plasma membrane, they become activated and thereby trapped within cells as their CoA derivatives. However, because the chain length specificity for activation versus uptake for each FATP is not completely identical, this mechanism alone cannot account for all of the experimental data. Furthermore, for FATPs not localized to the plasma membrane, this mechanism assumes that their ACS activity on internal cell membranes is sufficient to drive fatty acid uptake.
An alternative hypothesis is that of vectorial acylation, analogous to the situation in bacteria, where a specific outer membrane protein (FadL) and an intracellularly oriented acyl-CoA synthetase (FadD) are required for fatty acid uptake. This occurs in yeast, where the FATP homolog Fat1p and a longchain ACS (Faa1p or Faa4p) function together to facilitate fatty acid uptake (55). The finding that, in adipocytes, FATP1 and ACSL1 co-immunoprecipitated suggests that this might occur in mammals (34). DiRusso et al. (20) investigated this by expressing mouse FATPs individually in an S. cerevisiae mutant strain that was null for Fat1p and Faa1p and did not import the fluorescent fatty acid analog C 1 -BODIPY-C 12 . This compound is generally regarded as an analog of long-chain fatty acids and not VLCFA. Mouse ACSVL1, FATP1, and FATP4 were able to restore fatty acid uptake in this strain, whereas ACSVL2, ACSVL3, and ACSVL6 did not, even though all of the murine proteins were detected in the plasma membrane. Increased activation of C18:1, C20:4, and C24:0 was detected in all but the ACSVL6-expressing strain, suggesting a possible discordance between a transport-related role and the inherent ACS activity of the mouse FATPs. However, the precise mechanism by which an FATP functions in this model has not been elucidated.

Conclusion
Much progress has been made since the discovery of FATP in 1994 and the purification of the first enzyme with ACSVL activity in 1996. The members of the mammalian ACSVL/FATP family have all proven to be biochemically interesting and physiologically important proteins. Some of their proposed metabolic functions are shown in Fig. 1. Heart disease, obesity, diabetes/insulin resistance, cold intolerance, fat malabsorption, neurodegeneration, and defective epidermal barrier function are just a few conditions that may be associated with defective ACSVLs, yet many questions remain unanswered regarding the function, mechanism, and regulation of these enzymes. In particular, the mechanism by which ACSVL proteins facilitate cellular fatty acid uptake must be elucidated. Although metabolic trapping represents the simplest mechanism, it cannot completely explain all of the experimental data. There is precedence for vectorial acylation in bacteria and yeast, but how an ACSVL could function in this model must be elucidated. Similarly, if ACSVLs are truly bifunctional proteins, acting as ACSs under certain circumstances and as pure transporters under others, neither the transport mechanism nor the regulatory control mechanisms have yet been determined. This minireview has focused primarily on the function of the ACSVL enzymes as ACSs. Several recent reviews have addressed transport aspects of the FATP family in detail (31,36,41). A complete under-standing of ACSVL physiological function(s) awaits further investigation.